Kevin Rose

Let's get as much rich data as we can.

Let's do a very basic clustering algorithm even before we put in vector embeds.

What does that look like?

How does it feel?

And then I just was like, well, what is...

tech me not doing that I really care about.

And, and that led to my personal curiosity around novelty of, of objects.

How can I detect these trends before they become big?

You know, all these things where I was just like, I'm not seeing that anywhere else.

Um,

So I should just build it, right?

And so it's one feature at a time.

So what you're seeing here actually is like, I would say each of these things that you're seeing is probably a day or two of just being like, well, let's see if this will work and what it looks like.

And then kind of using in today's tools, what I would do is I would just use compound engineering on cloud code and do a workflow and say, this is the idea that I have.

Actually, I probably would, well, it depends on how big of idea.

If it's a minor feature tweak, I would just do it that way.

If it's something bigger that I needed to flesh out that had technology that I didn't know about,

what the right choice was, then I would use AI as more of a sparring partner on that side.

So I'll give you a great example.

There are probably 10 competing clustering algorithms that you can use for news.